The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses the run_command tool to execute php artisan commands on the host system. This capability allows the agent to interact directly with the project's framework tools and the underlying shell.
[INDIRECT_PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it incorporates untrusted strings from user prompts or project files into shell commands.
Ingestion points: User-supplied input or data from files used as arguments for Artisan commands (e.g., in php artisan make:controller ${userInput}).
Boundary markers: While the instructions suggest a strict alphanumeric allow-list, no specific delimiters or hard boundary markers are defined for the prompt interpolation.
Capability inventory: The skill allows execution of system commands via the run_command tool.
Sanitization: The instructions explicitly direct the agent to validate arguments, check the APP_ENV variable to prevent production data loss, and maintain an immutable denylist of destructive commands.

artisan-commands